Advanced topic 3, 2013
On the Management of Virtual Networks In order to deal with Internet ossification, network virtualization may be a good approach to address this problem by enabling the deployment of novel network architectures. Network virtualization environments(NVEs) can support multiple virtual networks running simultaneously in a single physical infrastructure without affecting the efficiency and reliability of network service, which means diverse network architectures can coexist in a single infrastructure to make the full use of its capability. In fact, virtual network is very complex and miscellaneous, which may lead to bad performance of network transmission. Even worse it man cause configuration error to paralyze the network. Therefore, management of NVEs is of great importance to guarantee the performance of the physical infrastructure and the stable service provided by the virtual networks. Advantages ● Energy Consumption is effective and reliable ● Easily Controlled and Inspected ● Provisioned as Needed ● Easily Relocated ● Disaster Recovery Influences ·Network Devices + Virtualization Support Commodity network devices with virtualization support are becoming popular in the marketplace such as routers supporting multiple virtual routing tables, Programmable switches, etc ·Business relationships The traditional ISPs now are divided into infrastructure providers (InPs) and service providers (SPs). InPs are mainly responsible for offering physical resources to service providers (SPs), SPs provide a variety of services that can be accessed by end users on demand through virtual network. Management model for network virtualization When managing the network virtualization environment , A virtual node is isolated in a physical one and operated by different SPs, Virtual links can be specified by SP operator manually or automatically. Virtual Links can configure VLANs, MPLS, LSPs, GRE tunnels and so on. When the physical node is selected: Step1: the SP operator requests a virtual node creation to the InP operator Step2: infrastructure manager instantiates the requested node Step3: the agent of the physical node receives a request from the infrastructure manager Step4: the agent contacts the hypervisor Step5: hypervisor conductss the requested action to create virtual node in the physical node Qualitative criteria of management of NVEs 1. Management target：A management target refers to a managed component of an NVE node management: manage the operation of virtual and physical nodes, i.e. initial creation& node migration link management: manage the configuration and operation of physical and virtual links, i.e. link isolation &flow scheduling. network management: manage the entire virtual network including those span multiple physical networks 2. Management functions：the ways of management of network virtualization projects Provisioning: Allows SPs to instantiate and use virtual networks. Monitoring: Supports several other management tasks, such as fault management and billing. Interfacing: Defines how management applications communicate with NVE resources and enabling interoperability. 3. Management approaches: how managers and agents are organized in the management of NVEs. Centralized and distributed management ·Centralized''' :Responsible for performing all management tasks ·Distributed :Multiple nodes cooperate to accomplish the management tasks Autonomic and policy-based management ·Autonomic : Reduce human intervention and allows dynamic adaption to changes in the network ·Policy-based ''': enforce isolation among virtual networks by controlling access permissions for each SP Network virtualization projects There are many projects worked on network virtualization for the requirement of management. 4WARD(VNet) is proposed to manage multiple virtual networks hosted in a shared infrastructure, which can provide discovery, embedding, and instantiation resources. AUTOI is designed to develop autonomic management solutions for future Internet to establish end-to-end service by distributed management system. FEDERICA is proposed to build a large-scale networking infrastructure, which can adapt to new Internet protocols and architectures. In this system, NOC performs all management actions and create slices and add virtual resources to slice to SPs. Under this way, SP can perform configurations on their slices without affecting others. ProtoGENI is is a deployed prototype of the Global Environment for Network Innovations (GENI). It’s proposed to create slices composed of silvers. In this system, silvers are a kind of virtual computing and networking resources. UCLP is designed to provide and control optical networks through multiple domains, which allows end users to build interdomain lightpaths by SOA. VNARMS is a automatic management system controlled by virtual network resource manager. In this system, it uses root-VN, firstlevel-VN, secondlevel-VN.ect to specify the quality of services. OpenFlow is a management system that A switch can be probably sliced and allocated to different users. One issue that FlowVisor has to deal with is managing isolation among multiple slices Comparing the above proposals: For both 4WARD and AUTOI, Single management is not responsible for overseeing the management of InP. For both FEDERICA and ProtoGENI, A centralized network performs all administrative Management actions in the infrastructure, such as resource discovery, provisioning, and user control Research perspectives 1. federations and SLA negotiations enables access to larger-scale infrastructures(eg. protoGENI) Future development': '''guaranteeing performance for participating entities through SLA. 2. interplay between InP management and SP management cooperation to avoid conflicts, ensure stability. Future development':' InP: maximize the utilization of infrastructures, revenue SP: want predictable performance for virtual networks. 3. management of virtualized cloud computing environments ·'Dynamic resource scaling:' Dynamically modifying a previous resource allocation Nowadays': provide elasticity at the virtual machine level. Future development: dynamic capacity adjustment of network resources (e.g.,bandwidth) requires further investigation. '''Application-aware resource provisioning: characteristics of the applications are commonly ignored Future development': '''Adaptive application-driven resource provisioning 'Energy management' a significant portion of the operational costs of the InP Solutions': consolidating virtual resources into a small number of physical devices. Future development: a good trade-off between energy consumption and applications' performance ''Data center network management '' address configuration, traffic management, and flow scheduling. Future development: short vs. long flows; flow scheduling, bandwidth allocation, and leveraging the inherent path diversity of data center networks 4. management of SDNs Software-defined networking provides a way to program network devices and manage their behaviors. SDNs enable separation of control and data planes and their implementation in software, which make it easy to control how network devices forward packet '''·Management abstractions: current solution to SDNs: require network operators to develop customized management packages using low-level instructions of a network operating system e.g., NOX Future development: some adequate interfaces, protocols,OMNI system ·'Interoperability and management API:' SDNs can be deployed using virtualized forwarding resources from different providers using a variety of network operating systems and implementations. Future development''': '''stresses a need of a widely accept of management API References ·Esteves.R and Grancille.L,” On the Management of Virtual Networks”, IEEE Communications Magazine, July, 2013, pp. 80-88.